Available Models

Crucible offers two production models and one experimental preview model. All models are accessible via the same completions endpoint.

crucible-1

The standard production model. Optimized for accuracy across a wide range of reasoning tasks. Recommended for most production workloads.

  • Context window: 128,000 tokens

  • Reasoning modes: Standard, Deep, Fast

  • Max output: 8,192 tokens

crucible-1-mini

A smaller, faster version of crucible-1. Reduced parameter count with minimal accuracy loss on structured tasks. Best for high-throughput pipelines where latency is a constraint.

  • Context window: 32,000 tokens

  • Reasoning modes: Standard, Fast

  • Max output: 4,096 tokens

crucible-2-preview

Early access to the next generation Crucible model. Significantly improved performance on long-document reasoning and multi-source synthesis. Not recommended for production use yet — breaking changes may occur.

  • Context window: 256,000 tokens

  • Reasoning modes: Standard, Deep, Fast

  • Max output: 16,384 tokens

  • Access: Allowlist only. Request access from your dashboard.

Create a free website with Framer, the website builder loved by startups, designers and agencies.